Robotic Process Automation Architectures and Processes for Hosting, Monitoring, and Retraining Machine Learning Models

1. A system, comprising: an artificial intelligence (AI) center comprising one or more physical servers, the one or more physical servers configured to: store an ML model in a container, and execute the stored ML model responsive to a call from an activity of a workflow of a robotic process automation (RPA) robot, the call sent by the RPA robot, the RPA workflow defining a controllable execution order over time and a relationship between a set of activities comprising the activity that calls the ML model by an expression of the activity, at least one activity of the set of activities configured to utilize at least one driver that facilitates interaction between the RPA robot and software of a respective computing system, wherein contents of the container comprising the ML model are encrypted or obfuscated, or both, and the ML model is encrypted or obfuscated, or both, and the ML model is retrained when performance of the ML model falls below a confidence threshold.

2. The system of claim 1, further comprising: a computing system comprising a low code ML model deployment and training application for RPA, the low code ML model deployment and training application configured to deploy ML packages in one click without coding by users.

3. The system of claim 2, wherein the low code ML model deployment and training application is configured to facilitate selection from a catalog of ML models and to deploy the selected ML models to the AI center as a target computing environment.

4. The system of claim 2, wherein the low code ML model deployment and training application is configured to display a graph of ML models that are connectable to one another.

5. The system of claim 4, wherein the ML models in the graph are connectable in series, in parallel, or both.

6. The system of claim 1, wherein the one or more physical servers are a single node cluster installation.

7. The system of claim 1, wherein the system is an air gap system that does not permit external communications outside of a facility.

8. The system of claim 1, wherein the one or more physical servers of the AI center are configured to: store a plurality of datasets for a plurality of ML models, each dataset comprising similar types of data in a logical or physical grouping; retrain an ML model of the plurality of ML models responsive to a training condition being met using a subset of the plurality of datasets specified in a training configuration for the ML model or manually upon request; and deploy the retrained ML model to be called by RPA robots.

9. The system of claim 8, wherein the one or more physical servers of the AI center are configured to: receive one or more performance measures for the ML model being retrained; and generate one or more scores for the one or more performance measures during the retraining.

10. The system of claim 9, wherein responsive to the one or more scores improving, the one or more physical servers of the AI center are configured to deploy the retrained ML model in place of a previous version of the ML model.

11. The system of claim 9, wherein responsive to the one or more scores improving, the one or more physical servers of the AI center are configured to deploy the retrained version of the ML model, use both the retrained version of the ML model and the previous version of the ML model, and choose a result with a highest confidence.

12. An artificial intelligence (AI) center comprising one or more physical servers, the one or more physical servers comprising: memory storing computer program instructions; and at least one processor configured to execute the computer program instructions, wherein the computer program instructions are configured to cause the at least one processor to: store a plurality of datasets for a plurality of machine learning (ML) models, each dataset comprising similar types of data in a logical or physical grouping, retrain an ML model of the plurality of ML models responsive to a training condition being met using a subset of the plurality of datasets specified in a training configuration for the ML model or manually upon request, wherein the ML model is retrained when performance of the ML model falls below a confidence threshold, receive one or more performance measures for the ML model being retrained, and generate one or more scores for the one or more performance measures during the retraining, and responsive to the one or more scores improving: deploy the retrained ML model in place of a previous version of the ML model or deploy the retrained version of the ML model, use both the retrained version of the ML model and the previous version of the ML model, and choose a result with a highest confidence, and execute the retrained version of the ML model responsive to a call from an activity of a workflow of a robotic process automation (RPA) robot, the call sent by the RPA robot, the RPA workflow defining a controllable execution order over time and a relationship between a set of activities comprising the activity that calls the ML model by an expression of the activity, at least one activity of the set of activities configured to utilize at least one driver that facilitates interaction between the RPA robot and software of a respective computing system.

13. The AI center of claim 12, wherein the retrained ML model is stored in a container, and contents of the container comprising the retrained ML model are encrypted or obfuscated, the retrained ML model is encrypted, or both.

14. The AI center of claim 12, wherein the one or more physical servers are a single node cluster installation.

15. The AI center of claim 12, wherein the AI center is an air gap system that does not permit external communications outside of a facility.

16. A computer-implemented method, comprising: storing an ML model in a container, by one or more physical servers of an artificial intelligence (AI) center; and executing the stored ML model, by the one or more physical servers of the AI center, upon a call from an activity of a workflow of a robotic process automation (RPA) robot, the call sent by the RPA robot, the RPA workflow defining a controllable execution order over time and a relationship between a set of activities comprising the activity that calls the ML model by an expression of the activity, at least one activity of the set of activities configured to utilize at least one driver that facilitates interaction between the RPA robot and software of a respective computing system, wherein contents of the container comprising the ML model are encrypted or obfuscated, or both, and the ML model is encrypted or obfuscated, or both, and the ML model is retrained when performance of the ML model falls below a confidence threshold.

17. The computer-implemented method of claim 16, further comprising: deploying ML packages in one click without coding by users, by a low code ML model deployment and training application; and facilitating selection from a catalog of ML models and deploying the selected ML models to the AI center as a target computing environment, by the low code ML model deployment and training application.

18. The computer-implemented method of claim 16, further comprising: displaying a graph of ML models that are connectable to one another, by a low code ML model deployment and training application, wherein the ML models in the graph are connectable in series, in parallel, or both.

19. The computer-implemented method of claim 16, further comprising: storing a plurality of datasets for a plurality of ML models, by the one or more physical servers of the AI center, each dataset comprising similar types of data in a logical or physical grouping; retraining an ML model of the plurality of ML models, by the one or more physical servers of the AI center, responsive to a training condition being met using a subset of the plurality of datasets specified in a training configuration for the ML model or manually upon request; receiving one or more performance measures for the ML model being retrained, by the one or more physical servers of the AI center; generating one or more scores for the one or more performance measures during the retraining, by the one or more physical servers of the AI center; and deploying the retrained ML model, by the one or more physical servers of the AI center, to be called by RPA robots.

20. The computer-implemented method of claim 19, wherein responsive to the one or more scores improving the method further comprises: deploying the retrained ML model in place of a previous version of the ML model, by the one or more physical servers of the AI center; or deploying the retrained version of the ML model, using both the retrained version of the ML model and the previous version of the ML model, and choosing a result with a highest confidence, by the one or more physical servers of the AI center.

21. The system of claim 1, wherein the ML model is available and made callable to the RPA robot by description, and results of the execution of the ML model are provided to the calling RPA robot rather than the ML model itself.